Following strong successes in all 3 projects of our present program, we have decided to increase the focus of our program project by building our new program on one of them and are seeking separate funding for the other two. In particular, we focus this program on segmentation and discrimination of the geometric structure of 3D objects in medical images by our principled and innovative methods working at multiple levels of scale and based on training image sets. The program consists of four projects and three cores. One of the projects focuses on development and extension of image analysis methods based on single and multiple object models at multiple levels of scale, with the middle levels of scale specifically designed to represent the figural aspects of objects. Because of the successes of our image analysis methodology, it is now time for three projects to pursue medical uses of such methods. Two of these three focus on the segmentation of structures in the brain and discrimination of disease from their geometric structure in relation to the diagnosis of mental illnesses. The other medical project continues our present program's focus on radiotherapy treatment planning. To support the evaluation of the image analysis methods both geometrically and in medical terms, we continue the core on biostatistics, and we add a core on the development of methodology for validation of image analysis methods. Finally, a facilities and administration core provides infrastructure for the other projects.